It's a text table from a genotyping array. 23andMe uses four columns (rsid, chromosome, position, two-letter genotype) and marks no-calls as --. AncestryDNA uses five, splitting the genotype into two allele columns, with 0 for a no-call. Both companies describe positions against GRCh37. The file covers a preselected set of markers (Ancestry says more than 700,000), not your whole genome, and the companies say it isn't for medical use.
What the file is
23andMe and AncestryDNA don't sequence your genome. They use a genotyping array (often called a chip), which tests a preselected list of positions where people commonly differ. The "raw data" download is a plain-text table of the results: one row per tested position.
- 23andMe: the help center describes a zipped text file whose name begins with "genome". Some computers unzip it automatically, so you may find a
.txtinstead. - AncestryDNA: the help center describes a download named
dna-data-(date).zipthat contains a.txtfile.
The text is readable, but the files have hundreds of thousands of rows. Most text editors can open them. If you look in a spreadsheet, don't save over the original.
The 23andMe layout
The file opens with comment lines starting with # that describe the file. Read them, since they're the quickest way to see what you have. Then come four tab-separated columns:
| Column | What it holds |
|---|---|
rsid | The marker's identifier, usually a dbSNP "rs" number |
chromosome | 1–22, X, Y or MT |
position | The coordinate on that chromosome |
genotype | Your two letters together, such as AG |
This order (ID, chromosome, position, combined genotype) is the one the bcftools manual uses in its 23andMe conversion example. The same documentation notes that a haploid call, such as one on a sex chromosome, can be a single letter. 23andMe's help center says it reports genotypes on the plus strand.
The AncestryDNA layout
AncestryDNA's file also opens with # comment lines, followed by a header row and five columns. Ancestry's help article says the first column is the rsID, the second the chromosome and the third the position, and "the last two columns contain your data". That is one allele per column (allele1 and allele2) instead of a combined pair. Other details from the same page:
- X and Y can appear as chromosomes 23 and 24. Open-source parsers such as the snps library also map 25 to the pseudoautosomal region and 26 to mitochondrial DNA.
- Insertions and deletions are written as I (extra DNA) and D (missing DNA).
- Data is reported on the forward strand.
Which reference build the positions use
Both companies describe their positions against GRCh37 (also called Build 37, and equivalent to UCSC's hg19). Ancestry says its data is reported on the forward strand of GRCh37. 23andMe says GRCh37 is its default, and that its Browse Raw Data feature can also show positions on GRCh38. Check the comment lines of your own file. If they name a build, that settles it.
The build matters because positions differ between GRCh37 and GRCh38. Any tool you upload to has to assume the right one. Most consumer tools expect these files in their original form, so don't edit positions by hand.
No-calls: "--" and "0"
Not every marker produces a confident result. According to 23andMe's help center, an uncalled marker "displays '--' instead of a two-letter genotype." In AncestryDNA files, a missing call shows as 0 in the allele columns, which is how open-source parsers such as snps treat it. A no-call means "no result here", not "normal" and not "absent".
Called results aren't guaranteed either. 23andMe states that only a subset of markers have been individually validated for accuracy, and that its raw data is "suitable only for informational use and not for medical, diagnostic, or other use."
Why this isn't your genome
Ancestry says its test looks at "more than 700,000 DNA markers". The same page puts human DNA at more than 3 billion letters, so the array covers only a few hundredths of one percent of positions. 23andMe's marker count depends on which chip version processed your sample, so check the row count of your own file rather than relying on a published figure.
The consequences are practical:
- A variant that isn't on the chip can't appear in the file, whatever your DNA contains.
- A missing variant in the file says nothing about whether you carry it.
- A rare variant is only tested if the chip was designed to include it.
Tools that "fill in" untested positions use statistics, not your actual DNA; genotype imputation explained covers how that works. If your question needs the whole picture, sequencing is a different product. The comparison page sets the options side by side.
Commercial option · affiliate link
Want the whole genome and every raw file?
Dante Labs sequences the whole genome at 30× and lets you download the raw BAM, FASTQ and VCF files.
Check Dante’s current offer →Our code GENOME may apply; codes may not combine with sitewide sales. Why this is an affiliate link.
Whichever test you use, a result that looks medically important needs a clinician or genetic counselor and usually a clinical-grade confirmatory test before you act on it.
How it differs from a VCF
A sequencing provider's VCF and an array file can both list genotypes, but they work differently:
| 23andMe / AncestryDNA text | VCF | |
|---|---|---|
| Header | A few # comment lines | Structured ## lines defining every field and contig |
| Alleles | Your letters only | Reference (REF) and alternate (ALT) alleles, with your genotype coded against them |
| Quality | None per row | QUAL, FILTER and per-sample quality fields |
| Coverage | Preselected markers | Whatever the variant caller found across the sequenced DNA |
| Samples | One person | One or many |
Converting an array file to VCF is possible but needs the matching reference genome, because the file doesn't say which allele is the reference. The bcftools manual's example for a 23andMe-style file is bcftools convert -c ID,CHROM,POS,AA -s SampleName -f 23andme-ref.fa --tsv2vcf 23andme.txt -o out.vcf.gz, where -f points to a GRCh37 FASTA whose chromosome names match the file. Most people never need to do this. Our VCF header guide explains what a real VCF contains.
Handling the file safely
Both companies are explicit that a downloaded copy is your responsibility. Ancestry says your downloaded copy "is no longer under the protection of our security measures." 23andMe says it "cannot secure Personal Information that you release on your own."
- Keep the original zip unchanged, and work on copies.
- Store it encrypted rather than in an open cloud folder or email attachment. See how to encrypt genome files.
- Before uploading anywhere, read the service's privacy policy and confirm you can delete your data, as Ancestry itself recommends.
- Expect results from other sites to differ. Ancestry notes that other companies use different algorithms and databases.
- Remember the file says something about your biological relatives too.
The genetic privacy checklist walks through the rest.
Check your file locally
The genome file inspector recognises 23andMe-style and Ancestry-style text files in your browser, without uploading them. It reads the first few megabytes, names the format, and reports a genome build if the file's header mentions one. Unzip the download first, since it reads .txt and .csv files, not .zip. It identifies files; it doesn't interpret your genotypes.
Common questions
What does '--' mean in my 23andMe raw data?
It's a no-call: 23andMe couldn't confidently determine your genotype at that marker. It doesn't mean you lack a variant.
Why does my AncestryDNA file list chromosomes 23 and 24?
Ancestry numbers the X and Y chromosomes as 23 and 24. Open-source parsers also read 25 as the pseudoautosomal region and 26 as mitochondrial DNA.
Which genome build is my raw data on?
Both companies describe their positions against GRCh37 (Build 37). 23andMe's Browse Raw Data feature can also show GRCh38, so check the comment lines in your own file.
Can I use my raw data file for medical decisions?
No. 23andMe says its raw data is for informational use only, and only some markers are individually validated. Take anything concerning to a clinician or genetic counselor for confirmatory clinical testing.
Is a raw data file the same as a VCF?
No. It lists your letters at preselected markers with no reference/alternate alleles or quality fields. Converting it to VCF requires the matching reference genome.
Keep going
Sources
Checked in October 2026 against 23andMe's help center (accessing and navigating raw data, “not determined” results, reference genome and strand), Ancestry's help center (downloading DNA data, DNA markers), the bcftools manual's TSV-to-VCF section, and the open-source snps parser for AncestryDNA no-call and chromosome codes 25/26. Left out: a 23andMe marker count (it varies by chip version and we found no current official figure), exact comment-line wording (it varies by file version), and any accuracy percentages.
- 23andMe: Accessing Your Raw Genetic Data
- 23andMe: Which Reference Genome and Strand Does 23andMe Use?
- 23andMe: What Does “Not Determined” Mean?
- Ancestry: Downloading DNA Data
- Ancestry: What are DNA Markers?
- bcftools manual: convert --tsv2vcf
Educational content, not medical advice. A genetic result is information, not a diagnosis; for anything medically important, talk to a qualified clinician or genetic counselor. Some outbound links are affiliate links; see the disclosure.